Published Jan 17, 2022

Matrix Profiles in Stumpy

Explore the transformative power of matrix profiles in diverse real-world applications, as Sean Law delves into the STUMPY Python library's capabilities in time series analysis, revealing how it uncovers critical patterns and anomalies in sectors like music, finance, and earthquake prediction.
Episode Highlights
Data Skeptic logo

Popular Clips

Episode Highlights

  • Musical Analysis

    Matrix profiling's versatility shines in musical analysis, where it identifies motifs and patterns in music data. illustrates this with a famous example comparing the bass lines of "Ice Ice Baby" by Vanilla Ice and "Under Pressure" by David Bowie and Queen. By mapping audio into mel frequency space, demonstrates how can reveal striking similarities between these songs 1.

    You input these two time series and it will actually show you where the two songs are very, very well matched.

    ---

    This capability highlights the tool's potential in diverse fields beyond music, such as climate data analysis 1.

       

    Finance Applications

    Matrix profiles have significant applications in finance, particularly in trading and anomaly detection. explains how is used to analyze trading data, such as daily closing prices, to make informed investment decisions 2. Additionally, the tool's ability to handle large datasets efficiently makes it ideal for financial applications where quick analysis is crucial 3.

    Some of the use cases that we've seen...is to look at trading data.

    ---

    This adaptability underscores Stumpy's value in solving nonobvious problems with innovative data solutions 3.

       

    Earthquake Predictions

    Matrix profiles also assist in predicting natural phenomena like earthquakes through pattern recognition. discusses how patterns extracted from earthquake data can provide insights into seismic activities over decades 4. This approach is not limited to earthquakes; it can also analyze pandemic-related time series data to uncover historical patterns.

    There are certain patterns that can potentially be extracted from earthquake data.

    ---

    Such applications demonstrate the tool's potential in various scientific and real-world scenarios 4.